Loading presentation...

Present Remotely

Send the link below via email or IM


Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.


SALAD-2013-Knowledge Discovery meets Linked APIs-JHoxha

No description

Julia Hoxha

on 24 July 2013

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of SALAD-2013-Knowledge Discovery meets Linked APIs-JHoxha

Knowledge Discovery
Linked APIs

Knowledge Discovery meets Linked APIs
Julia Hoxha*
Maria Maleshkova
Peter Korevaar

Karlsruhe Institute of Technology (KIT)
Present few suggestions for potential synergies

Propose predictive model to capture the rich relational structure of Linked APIs

Stimulate the interest of communities to explore novel approaches for mutual research

Julia Hoxha*, Maria Maleshkova, Peter Korevaar
Karlsruhe Institute of Technology (KIT)
Knowledge Discovery
Linked APIs
Integration of Web APIs, or data-providing services, with Linked Data.

Enable the communication of APIs at a semantic level, so that they can consume and produce Linked Data.
Linked APIs

Definition (Fayyad et al. 1996):
"Knowledge Discovery (KD) is the non-trivial process of identifying valid, novel, potentially useful, and ultimately understandable patterns in data"

Data mining is an important part of the KD field.
Knowledge Discovery
& Data Mining
Example GeoNames API

(1) How can the KDD methods be leveraged and pushed forward using contributions from Linked APIs?

(2) How can we tackle the problems and research questions of Linked APIs by applying KDD methods?
Questions to address:
Shift from item/page recommendation to Web API request recommendation
Linked APIs for KDD
Challenge: building semantic models to describe and deploy APIs
Construct the inputs/outputs worksheet from the invocation URLs [Taheriyan et al. ISWC, 2012]
Pattern recognition, structure learning, or group detection [Getoor et al. 2007]
KDD for Linked APIs
- Combines formalism of expressive knowledge representation with statistical approaches

- Perform probabilistic inference and learning on relational networks
Statistical Relational Learning (SRL)
- Explore the rich relational information in the usage data of APIs

- Capture the relational structure of API requests
Why and How to apply SRL
for Linked APIs?
Web API Requests Database
Mining Web API Requests
API Requests Network
Learn the probabilistic dependencies between the random variables
Predict the relationship R
Model as basis for cluster analysis:
Hidden Relational Model (HRM)
Relationship Prediction
Solving tasks like
Gain insights about the usage of the APIs
HRM: Practical Applications
Relational Model of API Requests
Pattern Mining
Federated Search
Recommender Systems
Enable semantic (association) pattern mining based on ontology as knowledge representation
Reach big datasets through APIs
Benchmarking with large-scale semantic techniques (matching, linkage, etc.)
Federated entity discovery over distributed Web APIs
Abundant data in structured format and in distributed sources for IR community
Tasks: data completeness, ranking, or information redundancy via on-the-fly entity consolidation techniques
Group APIs together, based on how they have been frequently requested by agents
Cluster Analysis
Based on past usage logs, explore which API makes more sense to request next given a specific coming request
based on APIs matching driven by usage behavior
Predictive models to facilitate the automation of Web API compositions
Generate active knowledge on which APIs to link together to create useful mashups
cluster assignments (or hidden states) of the objects decided by their attributes and their relations
hidden variables interpreted as cluster variables where similar API requests are grouped together
Semantic Models of Web APIs
Web API Usage Mining
(i) event detection and pattern discovery
(ii) frequent pattern analysis
(iii) statistical relational learning for relation prediction or clustering
KSRI Oberseminar
July 24, 2013
KDD methods
(pattern mining, clusterign, graph mining, classification, regression, etc.)
applied or needed in the domain of
Web Services

World of services on the Web is marked by the increasing use of Web APIs, or data-providing services.

Linked Data Cloud, largest public collection of structured data

Linked APIs
Integration of Web APIs with Linked Data
Full transcript